Curriculum and Gender Norms · that education plays an important role in changing/shaping social norms about gender roles. While previous studies have shown that social norms a ect

Curriculum and Gender Norms:

The Effect of Co-Education of Home Economics∗

Hiromi Hara† Núria Rodŕıguez-Planas‡

February 1, 2019

Abstract

This study examines the causal effect of educational policy on people’s attitudes to-ward traditional gender roles through the study of a 1989 national educational reformin Japan in which home economics became a co-educational subject, having previouslybeen taken only by girls. Using a regression discontinuity design and microdata froma Japanese time-use survey, we examined whether co-educational classes in home eco-nomics have increased husbands’ participation in domestic production. We found asharp increase in both the amount and the share of the typical husband’s householdproduction at the cutoff point (the cohort born in the 1977 academic year), indicatingthat husbands who had studied home economics for six years in junior and senior highschool later contributed more to household production than those who did not. Thisimplies that by providing both men and women with a more equitable view of genderrole divisions, co-education can lead to increased male participation in activities pre-viously seen as gendered. The study also validates the role of education in reshapingsocial norms.

Keywords: gender norms, educational reform, co-education, gender gaps in householdproduction and the labor marketJEL classification: J22, J24, I2

∗Very preliminary. Don’t quote without permission. Permission for the use of micro data from the Surveyon Time Use and Leisure Activities given by the Statistics Bureau of Japan. This research was supported bythe Japan Society for the Promotion of Science (Grants-in-Aid for Scientific Research (C) #16K03711, andthe Fund for the Promotion of Joint International Research #15KK0097). Editorial assistance was providedby Philip C. MacLellan.†Associate Professor, Department of Social and Family Economy, Faculty of Human Sciences and De-

sign, Japan Women’s University, 2-8-1, Mejirodai, Bunkyo-ku, Tokyo 112-8681, Japan. E-mail: [email protected].‡Associate Professor, Economics Department, City University of New York, Queens Col-

lege, 306-G Powdermaker Hall, 65-30 Kissena Blvd., Queens, New York 11367, USA. E-mail:[email protected].

1 Introduction

Although gender equality has become a worldwide concern, many countries continue to ex-

hibit a gender gap in the labor market. Labor economists have been trying to explore its

cause: some have focused on the supply side, such as the gender difference in the human

capital and the child penalty,1 and the others the demand side, that is, the labor market

discrimination and the issues inherent in the labor market system.2 Furthermore, recently,

some studies focusing on the effect of the social norm have been published.

While differences in labor incentives may explain some of this imbalance in home produc-

tion, economists recently have begun to draw upon ideas about identity from the sociology

and social psychology literatures to examine how social norms can also play an important role

in people’s decision making (Akerlof and Kranton [2000] and Akerlof and Kranton [2010]).

As Bertrand et al. [2015] note, “They [Akerlof and Kranton] define identity as a sense of

belonging to a social category, coupled with a view on how people in that category should

behave. They propose that identity influences economic outcomes because deviating from the

prescribed behavior is inherently costly.” Bertrand et al. [2015] apply this model to gender

identity, with men and women as the social categories, and they focus on the traditional

behavioral prescription that “a man should earn more than his wife.” They show empirically

that even in couples where the wife’s potential income is likely to exceed her husband’s, the

wife is less likely to be in the labor force and, moreover, she also earns less than her potential

if she does work. Additionally, even in couples where the wife earns more than the husband,

the wife spends more time on household chores. As these results contradict the theoretical

expectation (Becker [1985], Becker [1965]), this suggests that gender identity does play an

important role in household behavior.3

In Japan, although the average wages of men and women have been gradually converging,

1Bertrand et al. [2010].2Cortés and Pan [2018], Gicheva [2013], and Goldin [2014].3See Fortin [2005].

1

with the female-male ratio of average monthly scheduled wages of full-time workers rising

from 59.7% in 1990 to 72.2% in 2014,4 this trend has stabilized in recent years. In 2014,

Japan’s gender wage gap remained third highest among OECD countries after Korea and

Estonia.5 6 With such an imbalance in labor incentives, it should not be surprising that

Japanese men spend relatively little time in unpaid home production. At around one hour

per day, it is the second lowest among developed countries after Korea.7

As is well known, Japan is one of the countries which have well-defined social norms in

which the husband has traditionally worked outside the home while the wife has specialized

in domestic production, and this is one of the reasons why the Japanese women are less likely

to work at the market intensively and the Japanese husband’s share of housework is relatively

small.8 Further, Rodŕıguez-Planas and Tanaka [2018] show that gendered social norms affect

Japanese women’s decision to work, which implies that the observed gender imbalance in the

labor market in Japan is caused by social norms that affect the behavior of both Japanese

men and women. It is thus possible that a change in the social norm might trigger a change

in behavior of men as well as women. Conversely, we might also say that a change behavior

might require a change in social norms, and we suggest that education policy might be one

means of achieving this.

Turning to the Japanese education system, the government guidelines for education

(GGE) are the legally binding standard for the education curriculum throughout the na-

tion, not by region, and have undergone periodic revision since they were first instituted at

the end of World War II. If the GGE are revised, all Japaneses schools have to change their

curriculum according to them.

Home economics, the field of study to acquire skills and knowledge related to home

4Japanese Ministry of Health, Labour, and Welfare Basic Survey on Wage Structure.5OECD database.6Hara [2018], Kawaguchi [2005], Kawaguchi [2015], and Chiang and Ohtake [2014] study the gender wage

gap in Japan.7The Statistics Bureau of Japan.8Porterfield and Winkler [2007] show that among teenagers in the US, time spent on housework by girls

is longer than for boys, suggesting that these social norms exist at an early age.

2

production such as cooking, sewing, housing and family studies, is currently a required subject

for both boys and girls, however, it was a required subject only for girls in junior high and

senior high school for an extended period after World War II, thus inculcating through

education policy the social norm that men do not need to do housework because that is a

woman’s domain.9 In addition, most classes in Japan occur in the homeroom class, with

teachers moving around the school and students staying in one place; however, some classes

such as science, music, art, and home economics take place in a special subject classroom,

such as the science room. Home economics was highlighted not only as a special subject since

it was held in the home economics classroom, but also a gender segregated one, for while

girls were studying home economics, boys were also outside their homeroom studying another

subject in a different specialized classroom. In this manner, by creating home economics as

a geographically separate knowledge domain hidden from boys, the social norm that home

production was applicable only to girls was actively promoted through both educational

policy and practice.10

However, the 1989 revision of GGE stipulated that both boys and girls born after the

beginning of the 1977 school year (FY1977) on April 2, 197711 would be required to study

home economics in junior and senior high school. We can infer that this shift in educational

policy to six years of home economics, taken co-educationally, might have contributed to a

substantial change in male attitudes about traditional gender roles and social norms regarding

gender roles. Accordingly, in this study, we use the difference in the education of home

economics for boys born before FY1976 (the pre-1977 cohort) and after FY1977 (the post-

1977 cohort) to examine the effect of education on men’s participation in home production,

using a regression discontinuity design and micro data from a Japanese time use survey

9In elementary school, it is a required subject for both boys and girls even now and then.10This began to change with the 1978 revision of the GGE, which determined that junior high school boys

would also take home economics. However, the required number of credits for boys was much smaller thanthat for girls, and home economics classes were gender segregated, with boys and girls taught separately indifferent classrooms.

11The school year in Japan begins in April and is thus different from the calendar year. To clarify, wedefine school year T (FYT ) to include the cohort born between April 2nd in T and April 1st in T + 1.

3

(JTUS).12

Of its several contributions to the field, first and most importantly, this study shows

that education plays an important role in changing/shaping social norms about gender roles.

While previous studies have shown that social norms affect decision to work (Rodŕıguez-

Planas and Tanaka [2018]) and even allow people to deviate from economically rational

behavior (Bertrand et al. [2015]), this is the first study to date that investigates factors

that might be effective in weakening these social norms. Other contributions include our

use of JTUS microdata that provides detailed information on time use (explained in detail

in Section 4.1) and a natural experiment framework that exploits the discrete change in

Japanese national educational policy which together provide a precise yet comprehensive

examination of the effect of education on social norms.

The main result of this study is that men of the post-1977 cohort are more likely to par-

ticipate in home production than those of the pre-1977 cohort, suggesting that co-education

in home economics changed the attitudes of both men and women about gender roles. This

implies that education is quite important in both forming and changing social norms.

The structure of this paper is as follows: We review previous literatures in Section 2.

Section 3 explains the change in educational policy on home economics in Japan, and Section

4 describes the data used in this study as well as the identification strategy for the analysis.

The estimation results are reported in Section 6, and Section 7 concludes the paper.

2 Literature Survey

There are a number of existing studies on determinants of gender role division. Roughly

speaking, it is shown that 1) economic development, medical improvements, and technological

change (e.g., Doepke and Tertilt [2009], and Albanesi and Olivetti [2016]),13 2) the production

12The formal name of JTUS is the Survey on Time Use and Leisure Activities conducted by the StatisticsBureau of Japan.

13Doepke and Tertilt [2009] show that increase in the importance of human capital by technological changeimproves in the legal rights of married women, and Albanesi and Olivetti [2016] do that medical progress isessential for the joint rise in women’s labor force participation and fertility.

4

structure of the economy and incentive problems in the labor market (e.g., Olivetti and

Petrongolo [2016] and Albanesi and Olivetti [2009]),14 3) demographic factor (the sex ratio)

(Cardoso and Morin [2018]),15 and 4) beliefs and norms about the appropriate role of women

in society are main factors. Thus, the gender norm, the fourth one, is considered to be one

of the most important factors determining the gender role.

Some studies, including both ones using survey data and ones using experimental frame-

work, show the effect of gender norm on people’s decision making. For example, Fernandez

and Fogli [2009] and Bertrand et al. [2015] show empirically that the social norms/the culture

are significant in explaining women’s hours of work and domestic production, and fertility

outcomes, using survey data like the General Social Survey. Additionally, using the experi-

mental framework, Bursztyn et al. [2018] and Bursztyn et al. [2017] also find it.

Now, how do people form/transmit believes and values on the gender role? From a long-

term/historical perspective, Alesina et al. [2013] provide an answer on the origin of gender

norm, that is, differences in traditional farming practices (“plough use”) have shaped the

evolution of norms and beliefs about the gender role in society.

Then, in the modern society, what form/transmit the gender norm? There might be some

factors; from parents to children, from peer to peer, from third parties such as media, experts,

the state, and schools. This study explores determinants of the gender norm, focusing on the

role of education. To our knowledge, this is the first study to examine the effect of education

on the form of gender norm.16

14Olivetti and Petrongolo [2016] provide the evidence on the interplay between the rise of the serviceeconomy and the growth in female hours of work. Albanesi and Olivetti [2009] show that the fraction ofperformance pay in total compensation is inversely related to home hours.

15Cardoso and Morin [2018] find that as the ratio of men in the society declines, female labor forceparticipation increases, women enter traditionally male-dominated occupations and industries.

16In terms of the political attitude, it is shown that the curriculum change could affect it (Cantoni et al.[2017] and Chen et al. [2016]).

5

3 Home Economics Education in Japan

3.1 Japanese Education System

Japan has a national education system, with the two main laws regulating Japanese schooling

being the Basic Act on Education and the School Education Law. The former is the fun-

damental law concerning school education in Japan, while the latter determines the school

system. In addition, a school’s curriculum is also regulated by various other laws, govern-

ment ordinances, regulations, and notifications. Among them, the most important are the

Government Guidelines for Education (GGE),17 which are determined based on the School

Education Law.

The GGE, which are determined separately for elementary, junior high, senior high,

schools for the blind, schools for the deaf, and schools for the disabled, provide the legally

binding national curriculum standards for each type of school. As each school must adhere

to these guidelines on curriculum organization and course content, the GGE plays the most

important role in how education is practiced within the Japanese school system.

3.2 Home Economics Education before the 1989 Policy Change

In Japan, under the GGE, home economics has been studied in elementary, junior high and

senior high schools since the end of World War II, with the focus on learning content and

skills related to home production such as cooking, sewing, housing and nursing, as well as

understanding family, community and societal relationships.18 In elementary school, it has

been a required subject for both boys and girls throughout the post-war period, however,

it was not co-educational in junior high and senior high school before the policy change in

1989, shown in Table 1.

17Compulsory education in Japan begins at age six and consists of six years of elementary school and threeyears of junior high school. While not compulsory, most Japanese continue on to the three years of highschool. The high school enrollment rate has been greater than 95% since 1992, and in 2017, it was 96.4%(Ministry of Education, Culture, Sports, Science and Technology School Basic Survey).

18Yokoyama [1996] was referred to for this subsection.

6

In junior high school, the subject was originally in 1958 called “technology education

and home economics (gijutu-kateika).” Although treated officially as a single subject, in

practice, the content was divided by gender, with technology education for boys and home

economics for girls, and studied separately in different rooms. In Japan, most subjects such

as math, social studies and languages are held co-educationally in the students’ homeroom,

while subjects such as science that have special equipment needs require boys and girls to

go together to the science room for instruction. In contrast to this, for technology education

and home economics instruction, boys moved to the school workshop to study technology

while girls moved to the home economics room to study home economics.

This gendered implementation of the course began to change somewhat with the 1978

revision of the GGE which determined that junior high school boys would also take home

economics and girls would take technology. However, the required number of home economics

credits was substantially smaller for boys than for girls (and similarly for girls with respect to

technology), so the content coverage remained different according to gender.19 Additionally,

as boys and girls continued to study only with their own gender and not co-educationally,

the 1978 revision maintained the course as content differentiated and physically segregated

by gender.

3.3 Home Economics Education after the 1989 Policy Change

Although the gender segregation and content differentiation in the teaching of home eco-

nomics and technology had been a topic of discussion for some time in Japan from the

perspective of equality of educational opportunities for men and women, the 1989 policy

change was triggered by international pressure rather than by grass-roots movements based

on domestic public opinion.

19Technology education consisted of nine areas: wood-processing I and II, metal-processing I and II,machinery I and II, electricity I and II, and cultivation, while home economics consisted of eight areas:clothing I, II, and III; food I, II, and III; housing; and nursing. Junior high schools were then required tochoose five areas from technology education and one area from home economics for male students, and onearea from technology education and five areas from home economics for female students.

7

Table 1: Home Economics Education before Policy Change (–1989)

school subject name contents

junior high school technology education divided by gender(period of enrollment: and home economics • boys: mainly took technology

3 yrs, 13–15 yrs old) (large required # of technology credits& small # of home economics credits)

• girls: mainly took home economics(large required # of home economics credits

& small # of technology credits)

senior high school home economics • boys: not a required subject (no boys who took it)(period of enrollment: • girls: a required subject

3 yrs, 16–18 yrs old)

In December 1979, after six years of deliberation, the 34th General Assembly of the United

Nations adopted the Convention on the Elimination of All forms of Discrimination against

Women (CEDAW). In order to ratify the agreement, Japan needed to overcome some gender

inequality hurdles in three areas: nationality, employment, and education. With respect to

education, CEDAW required that all signatory states “shall take all appropriate measures

to eliminate discrimination against women in order to ensure to them equal rights with men

in the field of education,” including the same conditions for vocational training at all levels

of schooling, access to the same teaching staff and standard of curriculum quality, and the

elimination of stereotyped concepts of gender roles by revising textbooks, adapting teaching

methods, and promoting co-educational training (Article 10, (a) (b) (c)). Equal access to

vocational training was also required under the measures to eliminate discrimination against

women in the workplace (Article 11 (c)). While Japan desired to ratify CEDAW as a member

of the international community of developed nations, the home economics curriculum did not

satisfy these requirements, with gender-segregated and differentiated content in junior high

school and single-sex education for girls only in senior high school.

The discussion to revise the Government Guidelines for Education (GGE) thus began

in the Ministry of Education in 1984, and revisions for both junior and senior high schools

8

were published in March 1989 (hereafter, the 1989 GGE), eliminating the aforementioned

gender differences in home economics education in both junior and senior high school. At

that time, it was also decided that the new 1989 GGE would be adopted by junior high

schools in FY1993, but to ensure a smooth transition from the existing 1978 GGE to the

new 1989 GGE, a transition period was established between FY1989–1992 (the years that

cohorts born in 1976–1979 were enrolled in junior high school) that allowed schools flexibility

in deciding which year they would choose to adopt the new guidelines. In terms of home

economics education, special exceptions were set for the cohorts born in 1977–1979 (the 1977–

79 cohorts), but not for the 1976-cohort,20 meaning that the gendered difference in learning

areas in junior high school was still accepted for the 1976 cohort but were not permitted for

the 1977–1979 cohorts regardless of which GGE, 1977 or 1989, was in place at that school at

that time.21

Most importantly for our research, each junior high was not permitted to set any difference

between boys and girls by the 1989 GGE.22 Consequently, even though there may have been

slight differences in the content areas across junior high schools and the year in which each

school implemented the new 1989 GGE, for the post-1977 cohort, boys and girls in junior

high school studied the same content in the same classroom in a co-educational manner. For

this reason, the 1977 cohort is the cutoff of the regression discontinuity design in this study.

20When the 1976 cohort enrolled junior high schools, the 1977 GGE remained in place but junior highschools were required to consider the principle of the new 1989 GGE within the 1977 GGE framework. Inpractice, this meant that the gender differences in the learning areas for technology and home economicscontinued to be permitted.

21Looking in more detail at how the curriculum was changed by the 1989 GGE, in junior high school, tech-nology education and home economics now covered eleven content areas: wood processing, metal processing,electricity, machinery, cultivation, basic information, family life, food, clothing, housing, and nursing. Ofthese, wood-processing, electricity, family life, and food were required topics in every school, and to thesewere added at least three other areas from the list based on regional characteristics and the needs of eachschool and its students.

22The change in senior high schools is beyond the scope of this study. The co-education of home economicsin in senior high schools was stipulated by the new 1989 GGE too, and it was enacted in FY1994. However,regional differences existed in high school home economics education. For example, from FY1973, publicsenior high schools in Kyoto prefecture began co-education of home economics for boys and girls. This wasabolished in FY1985 as an institution, however, as a matter of fact, many senior high schools continued ituntil FY1994 (?).

9

4 Econometric Framework

4.1 Data

This study utilizes micro data from the Japanese time use survey (JTUS) conducted by

the Statistics Bureau of Japan (SBJ).23 The JTUS is the pre-eminent government statistical

database providing the most comprehensive and reliable data on daily patterns of time al-

location and on leisure activities in the country. The main topics covered are: 1) time use

over a single day, 2) participation in leisure activities during the past year, and 3) frequency

of participation in leisure activities during the past year. 24 The survey has been conducted

every five years in October since 1976, and the most recent 2016 survey is the ninth. We

have received permission from the SBJ to use the database for this study and, to examine

the effect of making home economics compulsory for boys, we utilized the most recent 2016

data.

The JTUS adopts a two-stage stratified sampling method, with the primary sampling

unit being the enemuration district (ED) of the Population Census of Japan, 25 and the

secondary sampling unit being the household. 26 All persons aged 10 and over in the sample

households are asked to respond to the survey, and foreigners living in Japan are included. 27

The JTUS is conducted by enumerators who proceed door-to-door, though an online response

also became possible for the 2016 survey. The enumerators deliver the questionnaires to each

23The formal name of the JTUS is the Survey on Time Use and Leisure Activities (syakai-seikatsu-kihon-chosa in Japanese).

24For time use during a single day, two questionnaires are used. Questionnaire A adopts a pre-codingmethod while Questionnaire B adopts an after-coding method designed to elucidate time use in more detail.

25The Census has been conducted every five years since 1920, with the most recent held in 2015.26First, the entire country is divided into the 47 prefectural regions and sample EDs are selected in each

region. Within each selected ED, households are selected from lists of households compiled by enumeratorsbefore the survey begins. For the 2016 survey, a total of about 7,300 sample EDs and 88,000 householdswithin those EDs were selected. Within each sample household, all people aged 10 and over are asked torespond to the survey. In 2016, this amounted to about 200,000 people.

27There are several groups of people who are explicitly excluded from the survey samples: 1) foreigndiplomatic and consular corps (including their families and other members of their party), 2) foreign militarypersonnel and civilian employees (including their families), 3) the Self-Defense Force personnel living inbarracks or vessels, 4) sentenced prisoners or persons in reformatories, 5) persons living in social welfarefacilities, 6) inpatients of hospitals or clinics, and 7) persons living on the water.

10

surveyed household, collect the completed questionnaires, and interview the households as

necessary.

The JTUS dataset has several advantages, one being that it surveys time use over a single

day in nominal terms, not by categories or ranges. This allows us to obtain the exact amounts.

Additionally the survey design creates a dataset that represents well the households in Japan.

The JTUS is a large-sample dataset with more than 80,000 household samples every survey

year (e.g. 88,000 in 2016) and for this study, we used the sample weights to weight back the

data to create an analysis sample representative of the target population.

4.2 Analysis Sample

In this section, we describe how the analysis sample of this study was constructed. Focusing

on those who were born 10 years before and after the cutoff point of FY1977 (that is,

from FY1967 to FY1987), the analysis sample was restricted to married individuals because

home production activities differ between those who are married and unmarried. For one,

responsibilities about family life are more crucial for those who are married (See subsection

5.1 for detail).

Our concern is whether there might be manipulation in the forcing variable. Checking to

ensure that there was no discontinuity in the marriage rate for males at the cutoff point, we

see in Figure 1 which shows the marriage rate for males by birth year that although there

was no sharp decrease at FY1977, the marriage rate did decline after FY1977, becoming

more pronounced toward FY1987. However, when we compare this with the 2015 Japanese

National Census, we see that while the level is different, there is no difference in the trend in

the JTUS and CPS. This indicates that the decline in the marriage rate after FY1977 is not

a discontinuity but merely reflects the lower marriage rate among younger men and implies

that there is no selection manipulation in the analysis sample.

Figure 2 shows the distribution of the forcing variable: the school (fiscal) year in which

the husband was born. We can see that while there is a slightly larger gap between FY1977

11

Figure 1: Marriage Rate for Males by Birth Year

0.1

.2.3

.4.5

.6.7

.8

1967 1970 1977 1980 1987Birth year

2016 JTUS 2015 Census

Source: 2016 JTUS and 2015 Census.

and FY1978, no discontinuity is observed at FY1977. Intuitively, this is a reasonable finding,

for in order for manipulation of our forcing variable to occur, it would require a couple to

change the timing of a pregnancy to accommodate an educational policy change some 13

years or more in the future. Moreover, as it would have been incredible to expect in 1976

or 1977 the policy change that would occur in 1989, we do not need to be concerned about

manipulation of the forcing variable in this study.

The descriptive statistics are summarized in Table 2, and we present both the totals and

the statistics for those born in a three year window before and after the cutoff date (1974–1976

and 1977–1980) in order to see if there is a difference between the two groups. We can see

that there are some differences in individual characteristics such as age, number of household

members older or younger than age 10 and working style. There are also some differences in

the wife’s characteristics. However, as these differences might depend on age, we conducted

an RD estimation for each covariate to ensure that they did not occur discontinuously. 28

28We used the following equation for each covariate (X): Xi = birthyeari+βPost1977i+�i. The notations

12

Table 2: Descriptive Statistics by Birth Year

Husband’s birth year1974–1976 1977–1980 Total

Husbandage 40.64 37.24 *** 38.95

(0.94) (1.24) (2.02)years of education 13.91 13.97 13.94

(2.24) (2.28) (2.26)work hour 49.13 49.21 49.17

(9.38) (9.21) (9.29)annual income 507.10 471.10 *** 489.10(million JPY) (230.20) (207.20) (219.70)child (yes/no) 0.62 0.77 *** 0.70

(0.49) (0.42) (0.46)no of children 0.95 1.31 *** 1.13

(0.90) (0.94) (0.94)working styleregular workers 0.83 0.85 0.84

(0.37) (0.36) (0.37)non-regular workers 0.03 0.03 0.03

(0.16) (0.18) (0.17)self employed 0.14 0.12 ** 0.13

(0.35) (0.32) (0.33)not working 0.01 0.01 0.01

(0.10) (0.08) (0.09)His wifeage 39.46 36.49 *** 37.98

(3.69) (3.76) (4.01)years of education 13.78 13.83 13.80

(1.77) (1.86) (1.82)work hour 32.75 32.98 32.86

(12.98) (13.59) (13.28)annual income 193.20 192.30 192.70(million JPY) (160.30) (152.70) (156.60)working styleregular workers 0.26 0.28 0.27

(0.44) (0.45) (0.45)non-regular workers 0.43 0.38 *** 0.40

(0.50) (0.48) (0.49)self employed 0.06 0.06 0.06

(0.25) (0.23) (0.24)not working 0.25 0.29 *** 0.27

(0.43) (0.45) (0.44)

Source: Statistics Bureau of Japan 2016 Survey on Time Use and Leisure Activities.Note: Standard deviation is in parenthesis. *** and ** indicates that t-test rejects the null hypothesis (whichboth values are the same) with 1% or 5% statistical significance respectively.

13

Figure 2: Distribution of the Forcing Variable

0.0

2.0

4.0

6.0

7

1967 1970 1977 1980 1987birth year

2016 JTUS 2015 CensusNote: Sample size of 2015 Census is 10,014,777, and that of 2016 JTUS is 28,076.

Source: 2016 JTUS and 2015 Census.

Table 3 shows the results. All covariates except for the number of household members

below 10 years old show with statistical significance that there is no difference at the cut off

point (See Figure A1), which implies that any difference in the individual characteristics by

age is the result of a continuous change. We can thus say that no covariate affects our RD

estimate of the share of home production within a couple.

4.3 Identification Strategy

The JTUS requires respondents to answer the length of time they spend in each of twenty

activities 29 in units of fifteen minutes. In other words, responses are provided only for activ-

ities on which the husband spent at least 15 minutes. For this study, we focused on activities

related to home production such as housework and childcare and so, for our purposes, the

are the same as those in Eq. (2).29These include sleep, personal up-keep, meals, commuting to and from school or work, work, school,

housework, caregiving, childcare, shopping, transportation (excluding commuting to and from school towork), TV, radio, newspaper, magazine, rest and relaxation, educational training, hobby, sports, volunteeractivity and social services, association, health care, and other.

14

Table 3: RD Estimates for Covariates (β̂)Panel A. Husband

(1) (2) (3) (4) (5) (6) (7) (8)age years of regular non-regular self-employed not working work hour annual

education worker worker income

Post1977i -0.008 0.039 -0.016 0.002 0.020 -0.006 -0.330 -3.583(0.019) (0.089) (0.015) (0.007) (0.013) (0.003) (0.383) (8.728)

N 10,187 10,073 10,162 10,162 10,162 10,185 9,437 10,032

Panel B. Husband’s wife

(1) (2) (3) (4) (5) (6) (7) (8)age years of regular non-regular self-employed not working work hour annual

education worker worker incomePost1977i 0.085 0.044 0.003 -0.015 -0.002 0.015 0.53 10.611

(0.142) (0.072) (0.017) (0.019) (0.009) (0.017) (0.632) (7.223)N 10,187 10,087 10,181 10,181 10,181 10,187 7,042 7,428

Source: Statistics Bureau of Japan 2016 Survey on Time Use and Leisure Activities.Note: The estimation equation is Xi = birthyeari + βPost1977i + �i, therefore, all equations are controlledfor birthyeari. Standard error is in parenthesis.

time allocated to home production thus includes the amount of time spent on housework,

caregiving to household members, childcare, shopping, and transportation.

The husband’s share of each activity is calculated as follows:

Shareji =a husband time spent on activity j

[a husband′s time spent on activity j] + [a wife′s time spent on activity j]. (1)

Our identification strategy is an RD design, and the basic estimation equation is a standard

RD model as follows:

Shareji = birthyeari + βPost1977i +X′iγ + �i, (2)

where Sharei indicates the husband’s share of household production within a couple, birthyeari

is the fiscal year when the husband was born, Post1977 is a dummy variable that takes the

value one if individual i was born after April 1977 (otherwise zero), and Xi is a set of indi-

vidual covariates. Our coefficient of interest is β̂.

15

5 Graphical Results

Before moving to our estimation results, in this section we first discuss the results of a

graphical analysis of the effects of the home economics policy change by applying the method

provided by Calonico et al. [2015] 30 to data from the FY1967–1987 cohorts, which is 10 years

before and after the cutoff point (FY1977).

5.1 Home Production Time

We begin by observing the results of the graphical analysis of home production time, which

we have defined as the the amount of time for housework, caregiving to household members,

childcare, grocery shopping, and transportation related to home production. Figure 4 shows

the actual and fitted profiles of home production time per day by weekday and weekend

according to the husband’s birth year (a unit is a minute).

Panel A shows the total home production time of a couple on weekdays and weekends,

and we can see that the total home production time increases over time as the husband’s

birth year approaches FY1987, but there are no discontinuous jumps at the cutoff point.

We can thus infer that co-education in home economics does not greatly affect the total

home production time of a couple. Next, looking at the results of the graphical analysis for

husbands and wives separately, we can see in Panel C that there is no effect on the wife’s

home production time but in Panel B we observe a sharp positive jump of about 20 minutes

in home production time for husbands on the weekend.

These results imply that the post-1977 cohort of husbands who studied home economics

in junior high school spend more time than the pre-1977 cohort on home production on

the weekend, but not on weekdays. This suggests that studying home economics seems to

30The dots represent the local sample means over non-overlapping bins under evenly spaced partitions.The number of bins is selected according to the mimicking variance method which is explicitly tailored toapproximate the underlying variability of the raw data and is thereby useful in depicting the data in adisciplined and objective way. The lines are polynomial regression curves estimated to flexibly approximatethe population conditional mean functions for the control and treated units over a restricted support of therunning variable. For more detail, see Calonico et al. [2014].

16

have a positive influence on men’s participation in home production, but as there is little

opportunity for most Japanese men to increase the time they spend on home production on

weekdays because of their long work hours, we observe this effect only on the weekend.

Next, we investigated whether there might be a difference in time devoted to the various

home production activities. For this analysis, we decomposed home production into its

individual activities but focused on housework and childcare because time devoted to the

other activities was quite small. Figure 4 shows only the weekend results because that is the

condition in which a jump in home production time was observed. The forcing variable is

again the husband’s birth year. In Panel A, which shows the results for a couple, we see a

negative jump in housework time at the cutoff point but a positive jump in childcare time.

As these two activities had opposite effects, this might explain why no jump was observed

above with respect to a couple’s total home production time.

Looking at husbands and wives individually (Panels B and C), we see no jump in house-

work but a sharp jump in childcare that is especially noticeable for husbands but also salient

for wives as well. One possible interpretation is that because couples often participate in

childcare together on the weekend, if the husband increases the time he devotes to childcare,

his wife does as well.

To check this hypothesis, we conducted the same analysis using the wife’s birth year as

the forcing variable. Figure 5 shows the results, and we see that if we use the wife’s birth

year as the forcing variable, we do not observe a jump in childcare time but a continuous

increase instead. This implies that while co-education in home economics does not directly

affect the behavior of wives, the time they devote to it increases as an indirect consequence

of their husband’s increased participation.

5.2 Husband and Wife Share of Home Production

Next, we observe the effects of co-education in home economics on the husband/wife share of

home production among couples. Here, because the total time allocated to home production

17

Figure 3: Home Production Time

300

400

500

600

700

1967 1970 1977 1980 1987husband's birth year

a. Weekday

300

400

500

600

700


b. Weekend

min

A. Couple (Husband and Wife)

Sample average within bin 4th order global polynomial

050

100

150

200


a. Weekday0

5010

015

020

0


b. Weekend

min

B. Husband


250

300

350

400

450


a. Weekday

250

300

350

400

450


b. Weekend

min

C. Wife


Source: 2016 JTUS.

18

Figure 4: Housework and Childcare Time on the Weekend

010

020

030

0


a. Housework

010

020

030

0


b. Childcare

min

A. Couple (Husband and Wife)


020

4060

8010

0


a. Housework0

2040

6080

100


b. Childcare

min

B. Husband


050

100

150

200

250


a. Housework

050

100

150

200

250


b. Childcare

min

C. Wife (weekend)


Source: 2016 JTUS.

19

Figure 5: Wife’s Housework and Childcare Time on the Weekend (fv=wife’s birth year)

050

100

150

200

250

1967 1970 1977 1980 1987wife's birth year

a. Housework

050

100

150

200

250

1967 1970 1977 1980 1987wife's birth year

b. Childcare

min

Wife (weekend)


Source: 2016 JTUS.

Note: The “fv” indicates the forcing variable.

might differ among couples, an analysis of absolute time devoted to it is inadequate. Instead,

we took the total time devoted to home production for each couple and examined how the

husband’s role in the household might have changed by investigating his Shareji as defined

in Eq. (1).

Figure 6 shows the actual and fitted husband share of home production time, with the

sample for this analysis being all couples. The dots in the figure represent the average share

of home production time according to the husband’s birth year, while the lines, as before,

represent fitted regressions from models with a quadratic birth year profile. While a jump at

the cutoff point can be observed for weekdays (Panel A), an even more pronounced jump of

around three points can be seen in the results for weekends (Panel B).

Finally, we checked to see whether there were any differences according to the type of cou-

ple: worker-worker, employee-employee (which excludes self-employed workers), and worker-

housewife. Figure 7 reports the results, and from Panel A, we find that when both husband

20

Figure 6: Husband’s Share of Home Production within a Couple (All couples)

510

1520

25

1967 1970 1977 1980 1987birth year

A. Weekday

510

1520

25

1967 1970 1977 1980 1987birth year

B. Weekend

%


Source: Statistics Bureau of Japan 2016 JTUS.

and wife are workers, a sharp jump can be observed on the weekend and a smaller jump on

weekdays. In Panel B, however, which excludes self-employment, we can see sharp jumps

for both weekdays and weekends, while in Panel C, which shows the results for a husband

worker/housewife couple, no jumps are observed either for weekdays or weekends. Taking

these results together, the effect of co-education in home economics is observed most clearly

in couples that face strict time constraints. Further analysis of each of these categories is an

issue for future study.

21

Figure 7: Husband’s Share of Home Production within a Couple by Couple Type

510

1520

2530

1967 1970 1977 1980 1987birth year

a. Weekday

510

1520

2530

1967 1970 1977 1980 1987birth year

b. Weekend

%

A. Worker-worker couple


510

1520

2530

1967 1970 1977 1980 1987birth year

a. Weekday5

1015

2025

30

1967 1970 1977 1980 1987birth year

b. Weekend

%

B. Employee-employee couple


510

1520

2530

1967 1970 1977 1980 1987birth year

a. Weekday

510

1520

2530

1967 1970 1977 1980 1987birth year

b. Weekend

%

C. Worker-housewife couple



22

6 Estimation Results

Having gained intuition from the graphical analysis, in this section we present and discuss

the estimation results of the RD regression of Equation (2) discussed in section 4.3.

Table 4 shows the estimation results for the home production time of a husband. For this

analysis, we began with a sample of all couples and then restricted the analysis to cohorts with

5, 3, 2 and 1-year windows around the cutoff point in order to check robustness of the results.

Panel A shows the results of the weekdays and we cannot see any statistically significant

results, so we focus here on Panel B which shows the results of the weekends. Column (6)

reports our baseline estimation for the 1967–87 birth cohort, comprising a ten-year window

around the cutoff point of FY1977. We can see that the husband’s home production time for

the post-1977 cohort is 21.468 minutes larger than for the pre-1977 cohort at a 1% level of

statistical significance. As we can also see the same results for the estimation with 5, 3, and

2-year windows around cutoff point (Columns (7), (8), and (9)), our estimation results are

therefore considered to be robust. This indicates that men who had studied home economics

in junior high school devoted more time to home production, particularly on the weekends,

than those who had not.

Next, Table 5 shows the estimation results for the share (%) of home production time

within a couple and reports the results of cohorts with ten (baseline), 5, 3, 2 and 1-year

windows around the cutoff point. Looking at the effect according to the days of the week

(Panel A), we see that there is a statistically significant difference on only the result with

ten year windows. However, if we see the results of the weekends (Panel B), the husband’s

share of home production for the post-1977 cohort is 2.046 points larger than for the pre-1977

cohort at a 1% level of statistical significance. The estimation results using shorter five, three

and two year windows around the cutoff point, respectively, are statistically significant, and

the coefficients are not different from that of Column (6). Due to the similarity in the value

of the coefficients for these three estimations, we can conclude that the results appear to be

23

Table 4: RD Estimate of Home Production Time (β̂ (coefficient of Post1977), Husband)Panel A. Weekdays (mins)

(1) (2) (3) (4) (5)1967–1987 1972–1982 1974–1980 1975–1979 1976–1978

10 yrs 5 yrs 3 yrs 2 yrs 1 yrPost1977 4.082 -0.116 2.285 -2.612 11.755

(3.561) (5.086) (6.611) (8.150) (10.235)N 9,215 5,268 3,395 2,430 1,511

Panel B. Weekends (mins)(6) (7) (8) (9) (10)

1967–1987 1972–1982 1974–1980 1975–1979 1976–197810 yrs 5 yrs 3 yrs 2 yrs 1 yr

Post1977 21.468*** 18.018*** 22.774*** 21.217** 16.462(4.843) (6.786) (8.683) (10.340) (14.060)

N 16,784 9,417 6,019 4,345 2,613

Source: Statistics Bureau of Japan 2016 JTUS.Note: Standard error is in parenthesis. ** and * indicates 5% and 10% statistical significance respectively.

reliable.

7 Conclusion

This study examined whether co-education in home economics, which has been compulsory

for both boys and girls in Japanese junior high schools since FY1989, has reduced the tra-

ditional gender gap in household production in Japan associated with long-standing social

norms. For the analysis, we used a regression discontinuity (RD) design and micro data from

a Japanese time use survey, and found a sharp increase at the cutoff point in the time hus-

bands devoted to home production, particularly to childcare. We also found a sharp increase

in the husband’s share of household production at the cutoff point, indicating that those

men who studied home economics during junior and senior high school later contributed a

larger share of household production compared to those who did not. This implies that co-

education in home economics might provide boys with a more equitable view of gender role

24

Table 5: RD Estimate of Home Production Share (β̂ (coefficient of Post1977), Husband)Panel A. Weekdays (%points)

(1) (2) (3) (4) (5)1967–1987 1972–1982 1974–1980 1975–1979 1976–1978

10 yrs 5 yrs 3 yrs 2 yrs 1 yrPost1977 1.713** 1.594 1.114 0.051 0.825

(0.723) (0.972) (1.258) (1.474) (1.903)N 8,894 5,083 3,289 2,348 1,463

Panel B. Weekends (%points)(6) (7) (8) (9) (10)

1967–1987 1972–1982 1974–1980 1975–1979 1976–197810 yrs 5 yrs 3 yrs 2 yrs 1 yr

Post1977 2.046*** 2.147** 2.594** 3.585** 2.730(0.727) (0.992) (1.247) (1.489) (2.017)

N 16,079 9,015 5,764 4,155 2,503

Source: Statistics Bureau of Japan 2016 JTUS.Note: Standard error is in parenthesis. ** and * indicates 5% and 10% statistical significance respectively.

divisions, which might encourage them to participate more actively in home production as

adults.

A limitation of the study is that we cannot identify the specific path that might lead

to this change in attitudes towards home production. As home economics classes teach not

only content that might raise students’ awareness about gender roles but also skills in home

production, it is possible that men’s participation in home production might have changed

due to skill accumulation rather than changes in attitudes. In other words, we cannot yet

identify precisely whether the effect of human capital development or changing social norms

is stronger. However, we do note that we observed larger effects for childcare than for house-

work, and because most of the skills related to childcare are considered to be acquired through

actual “on-the-job training” rather than school education, this might comprise supplemen-

tary evidence that co-education could change men’s attitudes and, eventually, the social

norms associated with gender roles. Although further study is required, the results of this

study show that education is crucial for changing social norms associated with traditional

25

household gender roles.

26

References

Akerlof, George A. and Rachel E. Kranton, “Economics and Identity,” The QuarterlyJournal of Economics, 2000, 115 (3), 715–753.

and , Identity Economics: How Our Identities Affect Our Work, Wages, and Well-Being, Princeton, NJ: Princeton University Press, 2010.

Albanesi, Stefania and Claudia Olivetti, “Home Production, Market Production andthe Gender Wage Gap: Incentives and Expectations,” Review of Economic Dynamics,2009, 12 (1), 80 – 107.

and , “Gender Roles and Medical Progress,” Journal of Political Economy, 2016, 124(3), 650–695.

Alesina, Alberto, Paola Giuliano, and Nathan Nunn, “On the Origins of GenderRoles: Women and the Plough,” The Quarterly Journal of Economics, 2013, 128 (2),469–530.

Becker, Gary S., “A Theory of the Allocation of Time,” The Economic Journal, 1965, 75(299), 493–517.

, “Human Capital, Effort, and the Sexual Division of Labor,” Journal of Labor Economics,1985, 3 (1, Part 2), S33–S58.

Bertrand, Marianne, Claudia Goldin, and Lawrence F. Katz, “Dynamics of theGender Gap for Young Professionals in the Financial and Corporate Sectors,” AmericanEconomic Journal: Applied Economics, 2010, 2 (3), 228–255.

, Emir Kamenica, and Jessica Pan, “Gender Identity and Relative Income withinHouseholds,” The Quarterly Journal of Economics, 2015, 130 (2), 571–614.

Bursztyn, Leonardo, Alessandra L González, and David Yanagizawa-Drott, “Mis-perceived Social Norms: Female Labor Force Participation in Saudi Arabia,” NBER Work-ing Paper No. 24736, 2018.

, Thomas Fujiwara, and Amanda Pallais, “’Acting Wife’: Marriage Market Incentivesand Labor Market Investments,” American Economic Review, 2017, 107 (11), 3288–3319.

Calonico, Sebastian, Matias D Cattaneo, and Rocio Titiunik, “Robust Data-drivenInference in the Regression-Discontinuity Design,” Stata Journal, 2014, 14 (4), 909–946.

, , and , “Optimal Data-Driven Regression Discontinuity Plots,” Journal of the Amer-ican Statistical Association, 2015, 110 (512), 1753–1769.

Cantoni, Davide, Yuyu Chen, David Y. Yang, Noam Yuchtman, and Y. JaneZhang, “Curriculum and Ideology,” Journal of Political Economy, 2017, 125 (2), 338–392.

27

Cardoso, Ana Rute and Louis-Philippe Morin, “Can Economic Pressure OvercomeSocial Norms? The Case of Female Labor Force Participation,” IZA Discussin Paper No.11822, 2018.

Chen, Wei-Lin, Ming-Jen Lin, and Tzu-Ting Yang, “Curriculum and Na-tional Identity: Evidence from the 1997 Textbook Reform in Taiwan,” mimeo,http://ir.sinica.edu.tw/handle/201000000A/79534, 2016.

Chiang, Hui-Yu and Fumio Ohtake, “Performance-Pay and the Gender Wage Gap inJapan,” Journal of The Japanese and International Economies, 2014, 34, 71 – 88.

Cortés, Patricia and Jessica Pan, “When Time Binds: Returns to Working Long Hoursand the Gender Wage Gap among the Highly Skilled,” Journal of Labor Economics, 2018.forthcoming.

Doepke, Matthias and Michele Tertilt, “Women’s Liberation: What’s in It for Men?,”The Quarterly Journal of Economics, 2009, 124 (4), 1541–1591.

Fernandez, Raquel and Alessandra Fogli, “Culture: An Empirical Investigation ofBeliefs, Work, and Fertility,” American Economic Journal: Macroeconomics, January 2009,1 (1), 146–77.

Fortin, Nicole M., “Gender Role Attitudes and the Labour-Market Outcomes of Womenacross OECD Countries,” Oxford Review of Economic Policy, 2005, 21 (3), 416–438.

Gicheva, Dora, “Working Long Hours and Early Career Outcomes in the High-End LaborMarket,” Journal of Labor Economics, 2013, 31 (4), 785–824.

Goldin, Claudia, “A Grand Gender Convergence: Its Last Chapter,” American EconomicReview, 2014, 104 (4), 1091–1119.

Hara, Hiromi, “The Gender Wage Gap across the Wage Distribution in Japan: Within-andBetween-Establishment Effects,” Labour Economics, 2018. forthcoming.

Kawaguchi, Akira, “The Analysis on the Determinants of the Gender Wage Differentialsin the 1990s,” Economic Analysis (Keizai Bunseki), 2005, 175, 52–82. (in Japanese).

, “Internal Labor Markets and Gender Inequality: Evidence from Japanese Micro Data,1990–2009,” Journal of the Japanese and International Economies, 2015, 38, 193–213.

Olivetti, Claudia and Barbara Petrongolo, “The Evolution of Gender Gaps in Indus-trialized Countries,” Annual Review of Economics, 2016, 8 (1), 405–434.

Porterfield, Shirley L and Anne E Winkler, “Teen Time Use and Parental Education:Evidence from the CPS, MTF, and ATUS,” Monthly Labor Review, 2007, 130, 37.

Rodŕıguez-Planas, Núria and Ryuichi Tanaka, “Gender Social Norms and Women’sDecision to Work,” Presented at the 30th Conference of the European Association of LabourEconomists, 2018.

28

Yokoyama, Fumino, “Transition of Educational Policy for Home Economics: from theGender Perspective in the Education Curriculum,” Hongo Journal of Law and Politics,1996, (5), 275–315. (in Japanese).

29

Appendix 1: Appendix Figures

Figure A1: Average Ratio of Husbands with Children under 10 Years Old and AverageNumber of Children under 10 Years Old by Husbands’ Birth Year

0.2

.4.6

.81

Rat

io

1967 1970 1977 1980 1987Husband's birth year

Panel A. Ratio of husbands w/ children under 10 yrs old0

.51

1.5

No

of c

hild

ren

1967 1970 1977 1980 1987Husband's birth year

Panel B. No of children under 10 yrs old per husband


30

Table A1: Estimation Results (Coefficient of treatment dummy: β)

(1) (2) (3) (4) (5)1967–87 1972–82 1974–80 1975–79 1976–7810 yrs 5 yrs 3 yrs 2 yrs 1 yr

Dependent variableA. Child dummy -0.004 0.009 -0.048 0.055 0(=1 if having children under 10 yrs old) (0.017) (0.027) (0.051) (0.034) (.)B. No of children under 10 yrs old 0.047 0.087 -0.044 0.141 0

(0.031) (0.054) (0.104) (0.070) (.)N 26,291 14,746 9,393 6,734 4,103

Notes:

1. SE is in parenthesis.

2. Yi = βPost1977i+γ1forcingi+γ2forcing2i +γ3Post1977iforcingi+γ4Post1977iforcing

2i +�i, forcingi

indicates distance from the cutoff point. The coefficient of our interest is β.

31

Table A2: Results of Housework Time on Weekend (β̂)Panel A. Husband

(1) (2) (3) (4) (5)1967–1987 1972–1982 1974–1980 1975–1979 1976–1978

10 yrs 5 yrs 3 yrs 2 yrs 1 yrPost1977 14.986 5.087 -1.217 -14.06 9.676

(6.565) (9.083) (11.472) (13.871) (18.798)N 17,730 9,973 6,380 4,604 2,772

Panel B. Wife(6) (7) (8) (9) (10)

1967–1987 1972–1982 1974–1980 1975–1979 1976–197810 yrs 5 yrs 3 yrs 2 yrs 1 yr

Post1977 -10.963** -14.995** -20.215** -21.252 -11.177(4.187) (5.623) (7.090) (8.480) (11.343)

N 17,730 9,973 6,380 4,604 2,772


Note: Shareji = α+ βPost1977i + γbirthyeari + �i. Standard error is in parenthesis. ** and * indicates 5%

and 10% statistical significance respectively.

Table A3: Results of Childcare Time on Weekend (β̂)Panel A. Husband

(1) (2) (3) (4) (5)1967–1987 1972–1982 1974–1980 1975–1979 1976–1978

10 yrs 5 yrs 3 yrs 2 yrs 1 yrPost1977 14.522*** 10.919** 15.027** 10.031 11.319

(2.881) (4.167) (5.478) (6.447) (8.801)N 17,730 9,973 6,380 4,604 2,772

Panel B. Wife(6) (7) (8) (9) (10)

1967–1987 1972–1982 1974–1980 1975–1979 1976–197810 yrs 5 yrs 3 yrs 2 yrs 1 yr

Post1977 23.326*** 18.377** 20.897* 10.435 20.656(4.460) (6.484) (8.335) (10.125) (14.123)

N 17,730 9,973 6,380 4,604 2,772


Note: Shareji = α + βPost1977i + γbirthyeari + �i. Standard error is in parenthesis. *** and ** indicates

1% and 5% statistical significance respectively.

32

Documents

Curriculum and Gender Norms · that education plays an important role in changing/shaping social norms about gender roles. While previous studies have shown that social norms a ect